Network Layer Utilities
The TCP/IP network layer uses several utility protocols to help it complete its task. For instance, in the first section of this chapter, you read that the Address Resolution Protocol (ARP) could be used to discover the MAC address of another IP host. In this section, you will learn about some basic IP utilities, using other protocols beside IP that together help IP deliver packets end to end through an IP network.
Address Resolution Protocol and the Domain Name System
Network designers should try to make using the network as simple as possible. At most, users might want to remember the name of another computer with which they want to communicate, such as remembering the name of a web site. They certainly do not want to remember the IP address, nor do they want to try to remember any MAC addresses! So, TCP/IP needs to have protocols that dynamically discover all the necessary information to allow communications, without the user knowing more than a name.
You might not even think that you need to know the name of another computer. For instance, when you open your browser, you probably have a default home page configured that the browser immediately downloads. You might not think of that URL string as a name, but the universal resource locator (URL) for the home page has a name embedded in it. For instance, in a URL such as http://www.skylinecomputer.com/Train_Welcome.asp, the http://www.skylinecomputer.com part is actually the name of the web server for the company that I work for. So, whether you type in the name of another networked computer or it is implied by what you see on your screen, the user typically identifies a remote computer by using a name.
So, TCP/IP needs a way to let a computer find the IP address of another computer based on its name. TCP/IP also needs a way to find MAC addresses associated with other computers on the same LAN subnet. Figure 5-7 outlines the problem.
)
Figure
5-7 Hannah Knows Jessie's Name, Needs IP Address and MAC Address
Hannah knows her own name, IP address, and MAC address because those things are configured in advance. What Hannah does not know are Jessie's IP and MAC addresses. To find the two missing facts, Hannah uses the Domain Name System (DNS) and the Address Resolution Protocol (ARP). Hannah knows the IP address of a DNS server because the address was preconfigured on Hannah's machine. Hannah now sends a DNS request to the DNS, asking for Jessie's IP address. The DNS replies with the address, 10.1.1.2. Figure 5-8 shows the simple process.
)
Figure
5-8 DNS Request and Reply
Hannah simply sends a DNS request to the server, supplying the name jessie, or jessie.skylinecomputer.com, and the DNS replies with the IP address (10.1.1.2, in this case). Effectively, the same thing happens when you surf the Internet and connect to any web site. Your PC somehow knows the IP address of the DNS; that information can be preconfigured or learned using Dynamic Host Configuration Protocol (DHCP), which is covered later in this chapter. Your PC sends a request, just like Hannah's request for Jessie, asking the DNS to resolve the name into an IP address. After that happens, your PC can start requesting that the web page be sent.
Back to the example with Hannah. Hannah still needs to know the Ethernet MAC address used by 10.1.1.2, so Hannah issues something called an ARP broadcast. An ARP broadcast is sent to a broadcast Ethernet address, so everyone on the LAN receives it. Because Jessie is on the LAN, Jessie receives the ARP broadcast. Because Jessie's IP address is 10.1.1.2 and the ARP broadcast is looking for the MAC address associated with 10.1.1.2, Jessie replies with her own MAC address. Figure 5-9 outlines the process.
)
Figure
5-9 Sample ARP Process
Now Hannah knows the destination IP and Ethernet addresses that she should use when sending frames to Jessie, and the packet in Figure 5-7 can be sent successfully.
ICMP Echo and the ping Command
IP needs to have a way to test basic IP connectivity, without relying on any applications to be working. Hannah, being a great network troubleshooter (in spite of being my 2-year-old daughter), can test basic network connectivity using the ping command. ping (Packet INternet Groper) uses the Internet Control Message Protocol (ICMP), sending a message called an ICMP echo request to another IP address. The computer with that IP address should reply with an ICMP echo reply. If that works, you successfully have tested the IP network. ICMP does not rely on any application, so it really just tests basic IP connectivity—Layers 1, 2, and 3 of the OSI model. Figure 5-10 outlines the basic process.
)
Figure
5-10 Sample Network, ping Command
ICMP contains many features, which are discussed in detail in Chapter 13, "Basic Router Configuration and Operation."
RARP, BOOTP, and DHCP
Over the years, three protocols have been popular to allow a host computer to discover the IP address it should use:
Reverse ARP (RARP)
Boot Protocol (BOOTP)
Dynamic Host Configuration Protocol (DHCP)
RARP and BOOTP work using the same basic process. To use either protocol, a PC needs a LAN interface card. The computer sends a LAN broadcast frame announcing its own MAC address and requests that someone assign it an IP address. Figure 5-11 outlines the process for both RARP and BOOTP.
)
Figure
5-11 RARP and BOOTP
RARP and BOOTP requests sent to the LAN broadcast address simply ask for an IP address assignment. Both protocols allow for IP address assignment, but that is all that RARP can ask for—it can't even ask for the subnet mask used on the LAN. RARP is defined in RFC 903, whereas BOOTP was defined later in RFC 1542, including several improvements over RARP. So, BOOTP allows many more tidbits of information to be announced to a BOOTP client—its IP address, its subnet mask, its default gateway IP addresses, its other server IP addresses, and the name of a file that the computer should download.
Both RARP and BOOTP were created with the motivation to allow a diskless workstation to come up and start operating. With RARP, the creators of the protocol just wanted to get the machine an IP address so that a knowledgeable user could type in commands and copy the correct files from a server onto the diskless computer's RAM memory so that they could be used. The creators of BOOTP, anticipating a less sophisticated user in the future, wanted to automate as much of the process as possible—including the dynamic assignment of a default gateway (router) IP address.
BOOTP's name really comes from the feature in which BOOTP supplies the name of a file to the BOOTP client. Typically, the diskless workstations had enough permanent memory to boot a very simple operating system, with the expectation that the computer would use a simple protocol, such as the Trivial File Transfer Protocol (TFTP), to transfer a file containing a more sophisticated operating system into RAM. So, with the ultimate goal being to let a diskless computer complete the processing of initializing, or booting, a full operating system, BOOTP was aptly named.
Neither RARP nor BOOTP is used much today. (They are possible topics for the INTRO exam, though.) One of the problems with both RARP and BOOTP is that they required a computer to act as a server, and the server was required to know the MAC address of every computer and the corresponding configuration parameters that each computer should be told. So, administration in a network of any size was painful.
DHCP, which is very popular in real networks today, solves some of the scaling and configuration issues with RARP and BOOTP, while supplying the same types of information. The main protocols for DHCP are defined in RFC 2131, but a couple of dozen additional RFCs define extensions and applications of DHCP for a variety of other useful purposes.
Like BOOTP, DHCP uses the concept of the client making a request and the server supplying the IP address to the client, plus other information such as the default gateway, subnet mask, DNS IP address, and other information. The biggest advantage of DHCP compared to BOOTP and RARP is that DHCP does not require that the DHCP server be configured with all MAC addresses of all clients. DHCP defines a process by which the server knows the IP subnet in which the DHCP client resides, and it can assign an IP address from a pool of valid IP addresses in that subnet. So, the DHCP server does not need to know the MAC address ahead of time. Also, most of the other information that DHCP might supply, such as the default router IP address, is the same for all hosts in the same subnet, so DHCP servers simply can configure information per subnet rather than per host and save a lot of administrative hassle compared to BOOTP.
The basic DHCP messages for acquiring an IP address are shown in Figure 5-12.
)
Figure
5-12 DHCP Messages to Acquire an IP Address
DHCP has become a very prolific protocol, with most end-user hosts on LANs in corporate networks getting their IP addresses and other basic configuration via DHCP.